Wikimedia says an AI copyright ruling could threaten the open web

In an amicus brief, the Wikimedia Foundation took neither side in copyright lawsuits involving publishers, OpenAI, and Microsoft. The organization argues that collecting copyrighted works to pretrain large language models (LLMs) may qualify as fair use when the result is new and different from the original sources, but that reproducing material in response to users should require a license when it substitutes for access to the original works.

Wikimedia says a nuance-free ruling could let copyright holders block transformative uses or give AI companies a green light to ignore licensing terms. The foundation also says these systems can help Wikipedia editors fight vandalism, copyright violations, and doxxing, even as services such as ChatGPT draw visitors away and may fail to follow Creative Commons license terms. The organization says an improper ruling could threaten the sustainability of free knowledge and the development of open-source technologies.